Skip to content

Sync failures

Cloud connectors can fail for very different reasons depending on the provider. This page catalogues the most common symptoms.

General diagnostic

# Connection status
podman exec myeline-web flask shell << 'EOF'
from app.models.cloud_connection import CloudConnection
for c in CloudConnection.query.all():
    print(f"{c.id} {c.provider} user={c.user_id} status={c.status} last_sync={c.last_sync_at} last_error={c.last_error}")
EOF

# Cron logs
podman exec myeline-cron cat /var/log/cron/check_cloud_sync.log | tail -50

# Force a manual sync
podman exec myeline-web flask run-cron check_cloud_sync

OAuth refused / redirect_uri_mismatch

Symptom: "OAuth connection error", redirect loop.

Cause: the redirect URI declared in your OAuth app (Google Console / Azure / Dropbox / etc.) doesn't match what Myeline generates.

Fix:

  1. Verify the URL generated:
podman exec myeline-web flask shell -c "
from flask import url_for
print(url_for('user_cloud.gdrive_callback', _external=True))
"

  1. Copy-paste this exact value into the "Authorised redirect URIs" list of the OAuth app on the provider side.

  2. Verify OAUTH_REDIRECT_BASE_URL in .env — must point at the exact public URL (e.g. https://myeline.acme.local), not a private IP.

Token expired and not refreshed

Symptom: last_error: "401 Unauthorized", status: error.

Cause: the refresh token was revoked (provider-side password changed, app deauthorised, inactivity period exceeded).

Fix: the user must reconnect their drive from /user/cloud ("Reconnect" button next to the connection in error). Sync history is preserved.

429 / Rate limit

Symptom: last_error: "429 Too Many Requests".

Cause: too many requests to the provider's API — typically on a multi-hundred-GB drive with forced sync.

Fix:

  • Wait — the sync resumes at the next cron (4 h).
  • Reduce frequency on the connection ("Sync interval" parameter on /user/cloud).
  • Make sure no other app is exhausting the same Google / Microsoft account's quota.

Files silently ignored

Symptom: a file is in the drive but doesn't appear in the library.

Possible causes:

  • Unsupported format (XLSX, PPTX, ZIP, video, image…). See the list of indexed formats in Library and uploads.
  • Size > 50 MB (silent reject indexer-side).
  • Provider-side encrypted document (Drive "Confidentiality protected", SharePoint "Information rights management"). The API refuses download.
  • Trash / recycle bin: trashed files are never indexed.
  • Folder filter: if the user picked a specific subfolder when connecting, files outside that folder are ignored.

Verification:

# Detailed worker logs during a sync
podman exec myeline-worker rq info
podman logs --tail 200 myeline-worker | grep -i "skip\|ignore\|reject"

Embedding failure

Symptom: the sync ends with success but the file stays at "In progress" indefinitely.

Cause: Ollama unreachable when the worker processed the file. The worker does 3 retries then gives up — the file is in the queue but not embedded.

Fix:

# Re-queue
podman exec myeline-worker rq requeue --queue myeline-pro --all
podman exec myeline-worker rq requeue --queue myeline --all

Then check Ollama state (see Ollama issues).

Standard tier — limits

Standard-tier accounts / licences have strict limits:

  • One active cloud connection only.
  • One sync per 24 h, regardless of cron cadence.

The /user/cloud progress banner explicitly shows the next allowed slot. No way to bypass without a plan / licence upgrade.